ArticlePDF Available

Automating Inequality: How High-Tech Tools Profile, Police, and Punish the Poor . By Virginia Eubanks. New York: St. Martin’s Press, 2017. Pp. 272. $26.99 (cloth).

Authors:
work, and send their children to school. Within the antiracism and anti-
oppression movements, there is an expectation that white people and seg-
regated white communities take action to interrupt racist patterns and
hold predominately white institutions accountable for past and current
injustice. Resources in this area include Showing Up for Racial Justice
(www.showingupforracialjustice.org) and Transforming White Privilege:
A 21st Century Leadership Capacity (www.racialequitytools.org).
Despite the editorsorientation toward collective harm and the collec-
tive benets of integration, most of the chapters still focus on the problem
of segregation for people of color,without recognition of the harm of seg-
regation for white and afuent communities. Many of the critiques of seg-
regation then advance a fairness argument about what white people should
do for people of color rather than an economic and social value case for
greater inclusion. Why not ask a different question: With equitable oppor-
tunities,what is the value that people of color can offer to communities and
to society as a whole?
Facing Segregation makes a purposeful call to action with straightfor-
ward policy solutions. Editors and chapter authors demonstrate that pol-
icy change is not only initiated at the state, regional, or federal level: Case
managers, property managers, and residents themselves can shift mind-
sets, practices, and systems. In the end, this volume contributes to both
the elds of social work and housing and community development by plac-
ing value on the multitude of actions neededfor advancing policy solutions
that seek to reverse trends of spatial inequities.
Amy T. Khare
Case Western Reserve University
Emily K. Miller
Case Western Reserve University
Automating Inequality: How High-Tech Tools Prole, Police, and Punish
the Poor. By Virginia Eubanks. New York: St. Martins Press, 2017. Pp. 272.
$26.99 (cloth).
Virginia Eubanks has written a terric book about the data and technology
systems that increasingly manage and control the lives of poor people in
the United States. The book will be of particular use to readers who are
Book Reviews |175
This content downloaded from 205.208.116.024 on May 08, 2020 05:39:30 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
largely unfamiliar with these issues, though it also offers plenty of inter-
esting insights for readers who already have signicant knowledge in this
area. After an introductory chapter that frames the book as a study of what
she calls the digital poorhouse,Eubanks takes readers through three care-
ful case studies. The astuteness of her approach lies in her recognition that
even the most seemingly benign implementations of data systems usher in
signicant social change likely to prove problematic for the poor individu-
als and families whose well-being is closely tied to how these data systems
function. Eubanks illustrates the many risks to which these kinds of data
systems expose not only the poor, but all of us.
Eubankssrst chapter introduces her framing device: the digital poor-
house. The chapter traces a history of the poorhouse in the United States,
followed by a recap of public welfare history since the New Deal. The an-
alytic point is to argue that the data systems described in later chapters rep-
resent signicant continuity in the nations approach to poverty manage-
ment and discipline. Social work and social policy professionals likely will
not nd much that is new in this history, but because the book appears
aimed at a wider readership, the chapter lays useful groundwork. The last
section of this chapter argues that although the physical poorhouse disap-
peared from US society decades ago, data-based approaches to govern-
ing poverty promise to build an ever tighter, always-on, evergreen digital
structure around the poor (always-onand evergreenare terms used
in data circles to describe the most desirable data systems: those that are
continually collecting new and up-to-date information, thereby enabling
the most real-time analysis).
The core of the book is Eubankss three case studies: the automation of
welfare eligibility determination by the state of Indiana, a centralized in-
take process for unhoused persons in Los Angeles, and an algorithmic risk
assessment procedure for the child welfare system of Allegheny County,
Pennsylvania. For each case, Eubanks tells powerful stories of a few indi-
viduals involved in the automated processas service users, workers, or
decision makers.Woven through these stories is the authors own insight-
ful commentary, explicating the larger issues that the individual stories il-
luminate. Each case paints a picture of a different dimension of the auto-
mated processes that increasingly govern what it means to be poor in this
country.
The tale of how Indiana automated its welfare eligibility system, and
the resulting human toll of what was sold as a routine exercise in increasing
|Social Service Review176
This content downloaded from 205.208.116.024 on May 08, 2020 05:39:30 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
government efciency, challenges us to recognize that technical reor-
ganization always comes with social consequences. In this case, those con-
sequences were severe, as illustrated by Eubankss account of the Stipes
family. The Stipess daughter, Sophie, was born in 2002 with severe disabil-
ities. Covered under Medicaid, the public health insurance program for
the poor, Sophie was able to access a range of services and supports. They
helped her make steady progress in growth and development, far more
than her doctors had anticipated at her birth. But in 2006, Indiana signed
a $1.16 billion contract with a group of companies led by IBM to privatize
the states eligibility determination system for Medicaid, Temporary As-
sistance for Needy Families, the Supplemental Nutrition Assistance Pro-
gram, and other state-run public assistance programs. Eubanks acknowl-
edges that the existing system was riddled with inefciencies and problems
but makes the case that the new, privatized system quickly began wreaking
havoc on Indianasmostvulnerablepeople.
The key problem was that most of the cost savings from automating el-
igibility processes came from large reductions in human labor. The most
consequential of these reductions was the laying off of caseworkers, who
worked from local ofces all around the state, could respond exibly to un-
usual situations, and generally believed thattheir job was to help the people
who came to them. Under automation, many of those caseworkers were re-
placed by data entry operators who could only respond to the binary choices
offered by a computerized system. Had the required client paperwork been
marked as received by the computer? No? Failure to cooperate,said
the system, unceremoniously cutting off eligibility. Never mind if the cli-
ents properly returned paperwork was mislaid before reaching a data en-
try operator, or if the operator inadvertently clicked the wrong button on
the screen, or if any number of other issues arose that a trained caseworker
with a relationship with the client might have resolved. At Amazon, such
errors might mean a customer has to place a new order for dog food. For
Sophie Stipes, lack of Medicaid coverage brought immediate, signicant,
and negative consequences for her health and well-being, and for that of
her worried parents. In the end,Eubanks writes, the Indiana automa-
tion experiment . . . denied [poor and working Americans] benets, due
process, dignity, and [in some cases] life itself(83). Although some might
point out that the welfare system had these effects well before the advent
of automation, the difference lies in the scale and impenetrability that an
automated data system enacts.
Book Reviews |177
This content downloaded from 205.208.116.024 on May 08, 2020 05:39:30 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
Eubankss second case study is from Los Angeles, where she tracks Los
Angeles Countys use of the Match.com of homeless services(84): a co-
ordinated entry database designed to more effectively connect unhoused
persons with the countys array of public and private services. As in Indi-
ana, the rationale for moving to an automated system lay in the previous
systems clear shortcomings. In this case, individuals were essentially forced
to compete with each other to secure any service that offered an opening,
regardless of whether it met their needs. The idea behind the coordinated
entry system is to obtain a birds-eye view of all clientsneeds as well as the
many service provider offerings and then to perform a dynamic, system-
wide sorting to deliver to more people the services from which they could
benet the most.
This approach, however, relies on what I would describe as a high-
resolution-data mind-set: the belief that people, places, processes, and
things can be known through the accumulation of many microbits of data,
which can then be sliced and diced mathematically to generate predic-
tions about the future.This is what Facebook, Instagram, Amazon, Google,
and all the other massive technology companies do: They collect every bit of
data they can about you, and then they serve you increasingly targeted sug-
gestions (i.e., advertisements) for products their algorithms predict you
will like and, hopefully, buy. The coordinated entry system in Los Angeles
thus relies on an intake survey driven by the high-resolution-data mind-
set: It asks for many, many pieces of data, much of it sensitive and intrusive.
The idea is that having all this information will improve the match of ser-
vices to your needs, a claim whose veracity Eubanks does not address,
though it poses an interesting empirical question. Eubanks does argue, how-
ever, that many unhoused persons in Los Angeles have completed the survey,
given up their sensitive data, and been offered no service at all. Needs on
Skid Row far outstrip the supply of services, and although those who are get-
ting services may indeed be better matched to the services that exist,what
about those who cannot access help? Not only are they still living on the
street, but now their sensitive information is contained in a database with
no specied time frame or procedure for removing it. With no regulation
of these data, who is to say what other entity, such as police, landlords, em-
ployers, or insurance companies, might nd a way to access it?
The nal case study of the book comes from Allegheny County, Penn-
sylvania. Its child welfare system is regarded by many local and national
observers as well run by sensitive leaders who care deeply about the children
|Social Service Review178
This content downloaded from 205.208.116.024 on May 08, 2020 05:39:30 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
and families in their care. One of the biggest challenges government child
welfare agencies face is accurately assessing what kind of intervention,
if any, is needed by a child about whose well-being there has been a re-
port made to the agency. In an effort to improve this assessment process,
over a period of years Allegheny County constructed a data warehouse con-
cerning individuals and families who utilized any of 29 different county or
state assistance programs. Caseworkers then searched these records when
determining whether to open an investigation following a call to the states
child abuse hotline. In 2012, partially motivated by budget cuts following
the Great Recession, the county human services agency was looking for ways
to improve this decision-making process. Although casework is its own type
of high-resolution data enterprise, it is often slow-moving and can be tainted
by personal bias. The county put out a request for proposals for data-based
decision tools, eventually selecting a predictive analytics approach designed
to improve its decisions on whether to open an investigation on a reported
case of abuse or neglect. Predictive analytics is a high-resolution data enter-
prise. Inside Allegheny Countys data warehouse lay a wealth of data, more
than a billion electronic records, from which to formulate automated predic-
tions of risk.
As of Eubankss writing, when Allegheny Countys intake screeners get
a report of a child at risk of abuse or neglect, they maketheir own risk assess-
ment based on the content of the report, whether it meets the legal stan-
dard for abuse or neglect, and what they learn about the child and family
from their manual inspection of the countys data warehouse. Only after
completing this process do the intake workers hit the button on the screen
to see what the algorithm says about the childs risk. Here a key difference
between the worker and the algorithm can have grave effect. The worker
sees the specic content of the reportfor example, a teenager sleeping in
a dirty, unheated house or a 6-year-old left outside by himself. But the al-
gorithm sees only the public-program utilization history of the child and
family (Medicaid enrollee, a parent on probation, public housing resident,
school attendance, etc.). Eubanks shows how a case in which the content
of the potential abuse report is fairly benign, but in which the family also
demonstrates substantial public-program involvement, is likely to get a
much higher risk score from the algorithm than a report of more severe
conditions in the home of a family with no public-program participation
history. In essence, the algorithm penalizes families who drew on govern-
ment support in the past. Eubankss recounting of the risk assessment
Book Reviews |179
This content downloaded from 205.208.116.024 on May 08, 2020 05:39:30 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
algorithm is chock full of fascinating, important detailthis review can-
not do it justice. It should be required reading for anyone working in child
welfare, probation and parole, or any other social service setting where
predictive analytics are being deployed under the claim that they will im-
prove decision making.
This is an excellent book on many levels: It is engaging, easy to read,
addresses an important topic, and clearly shows the authors expertise
and intellectual creativity around the subject. Data systems like the ones
Eubanks discusses have become pervasive in the lives of the poor, and it
is critical that social work and social policy professionals understand their
operational logic. The approach of this book reminds us that the mathe-
matical part of algorithms is the easiest piece to understand; it is the
choices that go into building the data systems that really matter and that
are most opaque. If anyone doubts the importance of developing a working
knowledge of these issues, consider what was, for a middle-class reader like
myself, one of the most arresting quotations in the book. It comes from a
young woman on welfare whom Eubanks interviewed years prior to under-
taking this project. The young woman told Eubanks that her caseworker
tracked her electronic benets card purchases and used what she found to
exercise discipline over her client.The young woman ended this story with
awarning:You should pay attention to what happens to us. Youre next(9).
Nicole P. Marwell
University of Chicago
Starving the Beast: Ronald Reagan and the Tax Cut Revolution. By Mon-
ica Prasad. New York: Russell Sage Foundation, 2018. Pp. 328. $35.95
(paper).
Over the past 40 years, the central mantra of the Republican Party has been
that taxes must be cut. Regardless of the issue or problem before them,
conservatives have consistently focused on reducing taxes as the answer
to our troubles.Yet those on the right side of the aisle have not always been
so eager for tax cuts. In fact, before 1980, Republicans were much more
concerned with balancing the budget and practicing scal responsibility.
What happened?
|Social Service Review180
This content downloaded from 205.208.116.024 on May 08, 2020 05:39:30 AM
All use subject to University of Chicago Press Terms and Conditions (http://www.journals.uchicago.edu/t-and-c).
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.